Tenant #1 is us. Tenant #2 could be you.
We route model selection for our own products through Maverick before asking anyone else to — every gap we find is a gap you never see. The engine is built and benchmarking nightly; now we're onboarding a small number of design partners.
- Onboarding on your real traffic — a working session mapping your workloads onto capabilities and tiers, not a sandbox tour.
- Policy setup done with you — pools, weights, and constraints configured against your actual compliance requirements.
- A say in what gets measured — design partners influence which model columns and capabilities ship next.
- Founding pricing — locked before public pricing exists.
Request access
Teams with real workloads to route get priority. Tell us what you're building.
No spam. No prompt ever stored — that policy starts with your email.
Three steps to a first recommendation.
We talk workloads
A short call about what you're routing today: which models, which tasks, what breaks, and what compliance boundaries apply.
We provision your tenant
API keys, pools, weights, and constraints — configured together so your first recommendation is already policy-correct.
You make one call
POST /v1/recommend from your app, or the MCP tools from
your agents. Ranked answer, check spec, decision trace. From there
it's just your gateway doing what it already does.